Neoadjuvant nivolumab with or without ipilimumab for cisplatin-ineligible patients with muscle-invasive bladder cancer

Abstract

Purpose:

Many patients with muscle-invasive bladder cancer (MIBC) are ineligible for cisplatin-based therapy. We conducted a phase II trial of neoadjuvant nivolumab±ipilimumab for cisplatin-ineligible patients.

Patients and Methods:

Patients with MIBC enrolled in two consecutive cohorts: (1) nivolumab alone; (2) ipilimumab/nivolumab. A third cohort with alternative dosing was planned. The primary endpoint was eligibility for cystectomy ≤60 days after last treatment. Correlative analyses were performed, with the PURE-01 trial used as an independent dataset.

Results:

Fifteen patients enrolled onto each cohort. In cohorts 1 and 2, 12/15 and 8/15 were eligible for cystectomy within 60 days, respectively. Due to cohort 2’s failure to meet the primary endpoint, cohort 3 was not initiated. With nivolumab alone, 4 patients achieved <ypT2ypN0 (26%), with 2 pathologic complete responses (pCRs) (13%). With ipilimumab/nivolumab, 3 achieved <ypT2ypN0 (20%), with 1 pCR (7%). One patient after nivolumab and 2 after ipilimumab/nivolumab had durable clinical CRs without cystectomy. 12-month recurrence-free survival (RFS) was 79% with nivolumab (95% confidence interval [CI], 61–100) and 61% with ipilimumab/nivolumab (95% CI, 39–95). Sequencing suggested NCOR1 alterations may associate with favorable outcomes. Gene expression profiling indicated potential association between tumor-infiltrating immune cells and longer RFS (log-rank p = 0.18); this was also observed in PURE-01 (p = 0.022).

Conclusions:

Among cisplatin-ineligible patients with MIBC, nivolumab alone was well tolerated. Ipilimumab/nivolumab caused toxicity that delayed cystectomy. Cases of progression before cystectomy indicated insufficient efficacy of pure neoadjuvant immunotherapy for unselected patients. Despite low response rates, some patients experienced sustained clinical CRs without cystectomy.

INTRODUCTION

Neoadjuvant cisplatin-based chemoimmunotherapy followed by radical cystectomy is a standard of care for muscle-invasive bladder cancer (MIBC) (1–3). Unfortunately, up to 50% of patients with MIBC are ineligible for cisplatin (4, 5), underscoring a need for novel neoadjuvant alternatives. Recently, perioperative enfortumab vedotin plus pembrolizumab demonstrated an overall survival benefit for cisplatin ineligible patients compared to cystectomy alone (6). However, a significant proportion of patients are also ineligible for enfortumab vedotin due to comorbidities such as peripheral neuropathy and hepatic impairment (7). Thus, additional neoadjuvant treatment options are still urgently needed.

In the last decade and a half, anti-PD(L)1 therapy has been established as a crucial therapeutic option for patients with bladder cancer in multiple settings (8–11). This has led to great interest in anti-PD(L)1 therapy alone or in combination with other immunotherapy agents, as a potential neoadjuvant strategy for cisplatin-ineligible MIBC. In some clinical contexts, addition of anti-CTLA4 inhibition to anti-PD(L)1 appears to improve therapeutic efficacy, as seen in clinical practice in advanced melanoma and renal cell cancer (12, 13).

Here, we reported the outcome of a multi-arm pilot study examining the safety of neoadjuvant anti-PD-1 immune checkpoint inhibition (nivolumab) with or without CTLA4 inhibition (ipilimumab) followed by radical cystectomy for cisplatin-ineligible patients with MIBC (ClinicalTrials.gov ID: NCT03520491).

PATIENTS AND METHODS

Patients

Eligible patients were men or women who were candidates for radical cystectomy plus pelvic lymph node dissection (RC-PLND) and had cT2N0M0-cT4aN0M0 disease as determined by cystoscopy and transurethral resection of bladder tumor (TURBT) within 60 days of enrollment as well as chest imaging and cross-sectional imaging of the abdomen and pelvis within 28 days of enrollment. Cisplatin ineligibility was defined as an estimated glomerular filtration rate < 60 mL/min per 1.73 m2 or pre-existing grade ≥ 2 peripheral neuropathy or hearing impairment (per CTCAE v4.0). Major exclusion criteria included prior use of immune checkpoint blockade for bladder cancer, autoimmune disease, New York Heart Association Grade II or greater congestive heart failure, and Karnofsky performance status less than 70.

Trial oversight

This study was approved by the institutional review board of Memorial Sloan Kettering Cancer Center and performed in accordance with the Declaration of Helsinki and Good Clinical Practice guidelines. All patients provided written informed consent before study entry.

Study procedures

This was a nonrandomized, open-label, sequential multi-cohort phase II study. The overall study was planned to include three sequential cohorts of 15 patients each treated with a unique regimen of immune checkpoint blockade. The regimens for each planned cohort were (cohort 1) nivolumab 3 mg/kg IV every 2 weeks for 5 doses; (cohort 2) ipilimumab 3 mg/kg plus nivolumab 1 mg/kg IV at weeks 0 and 6 with nivolumab 3 mg/kg IV alone at weeks 3 and 9; (cohort 3) ipilimumab 3 mg/kg plus nivolumab 1 mg/kg IV given on day 1 of a 21-day cycle for 3 cycles. RC-PLND was recommended within 60 days from the last dose of immunotherapy.

Immunohistochemistry

We performed PD-L1 immunohistochemical stain using a monoclonal antibody against PD-L1 (clone E1L3N, dilution 1:250; Cell Signaling Technology, Danvers, MA, USA; RRID:AB_2799389) on the Leica’s Bond III platform (Leica Biosystems, Buffalo Grove, IL, USA; RRID:SCR_026521). The staining condition of this antibody was optimized through internal validation and found comparable with PD-L1 22C3 pharmDx assay (RRID:AB_2833074). All slides were manually and independently reviewed by two genitourinary pathologists (HAA, JFC) and consensus was achieved in cases of initial discrepancy.

Sample collection and extraction

Based on pathological review, 5–24 unstained slides were macrodissected by scalpel or scraped by razor to collect tumor material in AutoLys M tubes (Thermofisher Life catalog # A38738). FFPE sections were digested with incubation in Protease Solution for 1 hour at 55°C followed by 1 hour at 90°C. DNA and then RNA were extracted sequentially using the MagMAX FFPE DNA/RNA Ultra Kit (ThermoFisher catalog # A31881) on the KingFisher Flex Purification System (ThermoFisher) according to the manufacturer’s protocol (RRID:SCR_008452). Samples were eluted in 50 μL (DNA) or 33 μL (RNA) elution solution. Genomic and transcriptomic analyses as detailed below were conducted in the Memorial Sloan Kettering Cancer Center Integrated Genomics Organization and Marie-Josée and Henry R. Kravis Center for Molecular Oncology core facilities.

Whole exome sequencing (WES)

After PicoGreen quantification and quality control by Agilent BioAnalyzer (RRID:SCR_018043), 100–200 ng of DNA were used to prepare libraries using the KAPA Hyper Prep Kit (Kapa Biosystems KK8504) with 8 cycles of PCR. After sample barcoding, 100–500 ng of library were captured by hybridization using the xGen Exome Research Panel v1.0 or v2.0 (IDT) according to the manufacturer’s protocol. PCR amplification of the post-capture libraries was carried out for 12 cycles. Samples were run on a HiSeq 4000 (RRID:SCR_016386) or NovaSeq 6000 (RRID:SCR_016387) in a PE100 run, using the HiSeq 3000/4000 SBS Kit or NovaSeq 6000 S1 or S4 Reagent Kit (200 Cycles) (Illumina). Normal and tumor samples were covered to an average of 93X and 172X, respectively. A subset of tumor samples targeted for deeper sequenced averaged 255X coverage.

WES sequencing data was analyzed using the TEMPO (Time-Efficient Mutational Profiling in Oncology; https://github.com/mskcc/tempo) pipeline. Briefly, demultiplexed FASTQ files were converted to BAM files and aligned to the b37 assembly of the human reference genome (14, 15). Somatic mutations were identified using MuTect2 (RRID:SCR_026692) (bioRxiv 2019.12.02.861054) and Strelka2 (16). Structural variants were identified using Manta (RRID:SCR_022997) (17) and Delly (RRID:SCR_004603) (18). Tumor mutational burden (TMB) was calculated as the number of non-synonymous exonic mutations per megabase of the target capture. MSI was assessed using MSIsensor (RRID:SCR_006418) (19): <3, microsatellite stable (MSS); ≥3 and <10, microsatellite indeterminate (MSI-I); and ≥10, MSI-high (MSI-H). Mutational signatures were inferred from single-nucleotide mutations for all sequenced samples with five or more somatic mutations. The fraction of mutations attributable to each of 30 known mutational signatures (20) was determined using a basin-hopping algorithm (https://github.com/mskcc/mutation-signatures), which assigns a weight to each of the 30 signatures based on the distribution of single-nucleotide substitutions and their trinucleotide context. Signatures with a known common source of somatic hypermutation were considered together e.g. signatures 6, 14, 15, 20, 21 and 26 as mismatch-repair deficiency/MSI-associated.

Transcriptome sequencing

After RiboGreen quantification and quality control by Agilent BioAnalyzer (RRID:SCR_018043), 0.1–1 μg of total RNA with DV200 percentages varying from 20–47% underwent ribosomal depletion and library preparation using the TruSeq Stranded Total RNA LT Kit (Illumina catalog # RS-122–1202) according to instructions provided by the manufacturer with 8 cycles of PCR. Samples were barcoded and run on a NovaSeq 6000 in a PE100 run, using the NovaSeq 6000 S2 or S4 Reagent Kit (200/300 Cycles) (Illumina; RRID:SCR_016387). On average, 65 million paired reads were generated per sample and 35% of the data mapped to the transcriptome.

Transcript abundance of baseline tumor samples was evaluated using Kallisto (RRID:SCR_016582) (21) aligned to GRCh37.75 human reference. A total of 10 baseline samples and 16 post treatment samples were transcriptionally profiled. To define sequencing and alignment quality we used CollectRnaSeqMetrics from Picard tools (RRID:SCR_006525; http://broadinstitute.github.io/picard/), and MultiQC v.1.9 (RRID:SCR_014982). Samples with low reads in the coding region and outliers using principal component analysis (PCA) were removed (n = 4). Differential gene expression was estimated using the R package DESeq2 (RRID:SCR_015687) (22). Genes that had an FDR-corrected p-value <0.05 and log2 fold change > 0 were used in Gene Set Enrichment Analysis (GSEA). h.all.v2023.1.Hs.symbols.gmt was used as the gene set database, Significant GO Terms (FDR q-value < 0.5) were ranked by the normalized enrichment score.

Immune cell deconvolution and immune subtyping

Transcript level abundances obtained by Kallisto (RRID:SCR_016582) were further aggregated to gene-level expression using sleuth R package (RRID:SCR_016883). The immune cell populations were quantified from variance stabilized RNAseq data using the immunedeconv R package (RRID:SCR_023869) (23) and its deconvolute function, along with the MCPcounter option. Immune subtypes were also identified using mibcCPIClass package (https://github.com/csgroen/mibcCPIclass) (24).

Comparison of PURE01 immune class with MSKCC cases

To compare PURE01 immune clusters with formerly derived clusters from the basal tumor samples we obtained centroids of the cell types from MCP-counter analysis from the case set. We then calculated Euclidian distance (distance = √Σ(Ai-Bi)2 ) between centroids of MSKCC cell types and each PURE01 cell type. Each sample was assigned immune hot or immune cold status based on the distance.

Cell-free DNA

Plasma cell-free DNA (cfDNA) sequencing was performed for cohort 2 using MSK-ACCESS, a cfDNA assay designed to detect mutations and select copy number alterations in 129 cancer-associated genes (25, 26). MSK-ACCESS uses unique molecular indexes and >15,000x depth of coverage that allow for an allele frequency detection threshold of 0.1% (26).

Oncogenic alterations were differentiated from variants of unknown significance using OncoKB (http://www.oncokb.org) (RRID:SCR_014782) (27).

Flow cytometry

Flow cytometry was performed in the Memorial Sloan Kettering Cancer Center Immune Monitoring core facility. Human peripheral blood mononuclear cell (PBMC) samples were thawed, washed, counted, and stained with a fixable Aqua viability dye (Invitrogen) and a cocktail of antibodies to the following surface markers: CD8-Qdot605 (Invitrogen, 3B5), CD4-Qdot 655 (Invitrogen, S3.5), PD-1-PE (BD, MIH4; RRID:AB_2869903), LAG-3-FITC (Enzo, 17B4; RRID:AB_10997322), ICOS-PE-Cy7 (eBioscience, ISA-3), TIM-3-APC (R&D Systems, 344823; RRID:AB_1964725). Cells were next fixed and permeabilized with the FoxP3/Ki-67 Fixation/Permeabilization Concentrate and Diluent (eBioscience). Cells were subsequently stained intracellularly with CD3-BV570 (Biolegend, UCHT1; RRID:AB_314063), Ki-67-AlexaFluor700 (BD, B56), FoxP3-eFluor450 (eBioscience, PCH101), and CTLA4-PerCP-eFluor710 (eBioscience, 14D3). Stained cells were acquired on a BD Biosciences LSRFortessa (RRID:SCR_018655) and analyzed using FlowJo software (FlowJo, LLC; RRID:SCR_008520). Isotype control stains were used for establishing positivity gates for PD-1, LAG-3, ICOS, TIM-3, FoxP3, and CTLA4.

Statistical analysis

The primary end point was safety of each cohort’s regimen, defined as ability to undergo RC-PLND within 60 days of the last dose of trial immunotherapy without delay due to treatment-related adverse events (TRAE) or disease progression. The trial aimed to test if the rate of proceeding to surgery without delay ≥60 days after immunotherapy was significantly higher than 60%. If at least 12 of 15 patients in a given cohort met these criteria, then that cohort would meet its primary endpoint and be considered safe. If the true rate of patients proceeding to surgery without delay ≥60 days was 88% (the rate observed in historical controls), this design had a power of 90% and type 1 error rate of 10%.

Secondary endpoints included pathologic downstaging to non-muscle invasive disease (<ypT2pN0) and complete pathologic response (ypT0pN0) rates, recurrence-free survival (RFS), and safety (Common Terminology Criteria for Adverse Events v5.0; RRID:SCR_010296). RFS was evaluated using the Kaplan-Meier method and measured from treatment initiation until disease recurrence, which was defined as investigator-determined clinical or radiographic progression. Event-free survival (EFS) was an exploratory endpoint. For RFS and EFS, patients without documented recurrence or death were censored at the time of last cross-sectional imaging. In addition, for purposes of RFS, patients who could not undergo RC-PLND due to TRAES were censored at time of last study treatment. For overall survival, patients without documented death were censored at time of last follow-up.

For purposes of pathologic response, patients who developed progressive and/or metastatic disease while on neoadjuvant therapy, those who were unable or unwilling to undergo RC-PLND, and those who discontinued protocol therapy because of treatment-related delays and/or treatment-related toxicity were considered pathologic non-responders.

Data availability

Cell-free DNA sequencing data are included in supplementary materials. Other data are available in dbGaP under accession phs001783. Additional data is available from the corresponding author upon reasonable request.

RESULTS

Between August 2018 to May 2021, a total of 30 patients were enrolled (Cohort 1, n=15; Cohort 2, n=15). Patient baseline characteristics are shown in Table 1. The representativeness of study participants is shown in Supplemental Table S1. The median Charlson comorbidity score of this cisplatin-ineligible cohort was 1, with a range of 0 to 5.

Table 1.

Baseline patient and disease characteristics by treatment cohort.

	Cohort 1 - Nivolumab alone (n = 15)	Cohort 2 - Ipilimumab + Nivolumab (n = 15)
Age (median, interquartile range)	75 (72, 78)	77 (65, 79)
Female, n (%)	4 (27%)	2 (13%)
White race	11 (73%)	13 (87%)
Hispanic ethnicity	0 (0%)	0 (0%)
Charlson Comorbidity Index (mean, range)	1.5 (0–5)	1.1 (0–3)
Baseline Karnofsky performance status, n (%)
100	0	2 (13%)
90	5 (33%)	11 (74%)
80	10 (67%)	2 (13%)
Creatinine in ml/min as per CKD-EPI, n (%)
≥60	5 (33%)	6 (40%)
45–59	5 (33%)	2 (13%)
<45	5 (33%)	7 (47%)
Prior intravesical BCG	4 (27%)	4 (27%)
Clinical T Stage, n (%)
cT2N0	8 (54%)	7 (46%)
cT3N0	7 (46%)	4 (27%)
cT4aN0	0	4 (27%)

Treatment and safety

Treatment details are listed in Table 2. Fourteen of 15 patients (93%) in Cohort 1 and six of 15 patients (40%) in Cohort 2 received all planned treatment. Seven of 15 patients from Cohort 2 received ≤ 50% of planned doses, all due to immune-related toxicities.

Table 2.

Treatment Details

	Cohort 1 Nivolumab alone (n = 15)	Cohort 2 Ipilimumab +Nivolumab (n = 15)
Received all planned doses	14 (93%)	6 (40%)
Received ≤50% of doses	1 (7%)	7 (47%)
Days from last treatment to RC-PLND (median, range)	41 (28–109)	75 (27–193)
Did not undergo RC-PLND, due to:	4 (27%)	6 (40%)
TRAE	1	1
Progression of disease	2	3
AE (not TRAE)	0	1
Declined surgery	1	1
RC-PLND delayed by TRAE >60 days after last treatment	0 (0%)	1 (7%)
Did not meet primary endpoint	3 (20%)	6 (40%)

AE, adverse events; RC-PLND, radical cystectomy with pelvic lymph node dissection; TRAE, treatment-related adverse events

In Cohort 1, eleven patients (73%) underwent radical cystectomy. Four did not undergo surgery due to progression of disease (n=2), TRAE (n=1, subsequently received bladder radiation) and patient refusal (n=1). Overall, 12 of 15 patients in Cohort 1 met the primary endpoint of ability to undergo RC-PLND within 60 days of the last dose of trial immunotherapy (though one of these patients declined RC-PLND despite ability to proceed within the specified time window). For Cohort 2, nine patients (60%) underwent surgery. Six did not undergo surgery due to progression of disease (n=3), adverse events (n=2, only one of which was TRAE), and opted for radiation (n=1), with 8 of 15 patients meeting the primary endpoint. The trial did not proceed to Cohort 3 due to Cohort 2’s failure to meet the threshold.

In Cohort 1, the only observed grade 3–4 TRAE was myocarditis (Table 3). In Cohort 2, 4 patients (27%) had grade 3–4 TRAE, including elevated lipase or amylase (n=3), pneumonitis (n=2), and transaminitis (n=2). Note, some patients experienced multiple grade 3–4 TRAEs.

Table 3.

Adverse Events

	Cohort 1 – Nivolumab alone (n = 15)		Cohort 2 – Ipilimumab + Nivolumab (n = 15)
	Any Grade n (%)	Grade ≥3 n (%)	Any Grade n (%)	Grade ≥3 n (%)
No. of patients with AEs of any cause	15 (100)	5 (33)	15 (100)	6 (40)
No. of patients experiencing TRAEs	12 (80)	1 (7)	14 (93)	4 (27)
TRAEs occurring in ≥3 patients or grade ≥3
Pruritus	4 (27)	0	8 (53)	0
Arthralgia	3 (20)	0	1 (7)	0
Lipase increased	3 (20)	0	4 (27)	3 (20)
Rash maculo-papular	3 (20)	0	5 (33)	0
Serum amylase increased	3 (20)	0	7 (47)	1 (7)
Aspartate aminotransferase increased	2 (13)	0	8 (53)	3 (20)
Diarrhea	2 (13)	0	4 (27)	1 (7)
Myocarditis	1 (7)	1 (7)	0	0
Alanine aminotransferase increased	1 (7)	0	6 (40)	2 (13)
Pneumonitis	0	0	3 (20)	2 (13)
Hypothyroidism	0	0	3 (20)	0
Hyperthyroidism	0	0	4 (27)	0
Blood bilirubin increased	0	0	2 (13)	1 (7)
Anemia	0	0	1 (7)	1 (7)
Constipation	0	0	3 (20)	0

AE, adverse events; No., number; TRAE, treatment-related adverse events

Anti-tumor activity

In Cohort 1, four patients had pathologic downstaging (26%), two of which had pathologic complete response (13%). In Cohort 2, three patients had pathologic downstaging (20%), one of which had a pathologic complete response (7%). Collectively, we observed a pathologic downstaging rate of 23% and pathologic complete response rate of 10%.

In Cohort 1, the patient who did not undergo radical cystectomy due to TRAE underwent bladder radiation and had clinical complete response but died of causes unrelated to cancer at 16.1 months. Two patients in Cohort 2 who did not undergo radical cystectomy had ongoing clinical complete response without additional treatment at 16.1 and 10.8 months of follow-up.

With median follow-up of 34.4 months for Cohort 1 and 19.1 months for Cohort 2, RFS at 12 months was 79% (95% CI, 61–100) and 63% (95% CI, 39–95), respectively. EFS at 12 months was 79% for Cohort 1 (95% CI, 61–100) and 66% for Cohort 2 (42–95). Overall survival at 24 months was 84% in Cohort 1 (95% CI, 66–100) and 66% in Cohort 2 (46–95) (Figures 1A-C).

Figure 1.

Survival outcomes, including (A) recurrence-free survival, (B) event-free survival and (C) overall survival. Cohort 1 received nivolumab monotherapy while Cohort 2 received nivolumab plus ipilimumab, both to be followed by radical cystectomy plus pelvic lymph node dissection. For purposes of brevity and exploration, both cohorts are depicted on the same Kaplan-Meier curves. However, the study was not designed to formally compare treatment arms, so no formal statistical comparison was performed.

Correlation of response with PD-L1 expression

Nineteen patients were evaluable for pre-treatment PD-L1 expression by IHC (Cohort 1, n=10; Cohort 2, n=9). In Cohort 1, six of 10 patients had PD-L1 expression on ≥1% of tumor cells, one of which experienced pathologic downstaging and one had durable clinical response with addition of radiation in lieu of cystectomy. One patient had PD-L1 expression on ≥5% tumor cells but was a pathologic non-responder. In Cohort 2, five of nine evaluable patients had baseline PD-L1 expression on ≥1% of tumor cells, of which one (20%) had pathologic (<ypT2N0) and one (20%) had clinical response without cystectomy; one of three patients (33%) with PD-L1 on ≥5% tumor cells experienced durable clinical response without cystectomy, while the other two were pathologic non-responders.

Genomic features assessed by whole exome sequencing (WES)

WES was performed on tumors from 14 patients from each cohort (Figure 2A-B). Median TMB was 6.3 mutations/megabase and all tumors were microsatellite stable (MSS). Fourteen percent (n = 4) patients had an oncogenic alteration in ERCC2. Previous reports have indicated an association between loss of function alterations in the nucleotide excision repair gene ERCC2 and sensitivity to cisplatin (28, 29), and ERCC2 along with other DNA damage response genes have also been identified as potential correlates of response to immune checkpoint blockade (30). All four patients with loss of function alterations ERCC2 had either pathologic downstaging or a durable complete clinical response (Figure 2A). A previous report also identified the chromatin modifying gene NCOR1 as a potential predictor of benefit from immune checkpoint blockade in bladder cancer (31). Three patients had truncating alterations of the chromatin modifying gene NCOR1, all of whom had pathologic or clinical response to ipilimumab plus nivolumab. None of the three patients with an NCOR1 alteration had disease recurrence or death after 10.5, 17.0, and 26.7 months of follow-up each (Supplemental Figure S1). However, association of NCOR1 with RFS was not significant after adjustment for TMB (p = 0.8).

Figure 2.

(A) Genomic characteristics of pre-treatment samples based on whole exome sequencing, with prevalence of oncogenic mutations per patient, grouped by treatment cohort and response. (B) Genomic characteristics by treatment response, pooling across treatment cohorts.

AMP, amplification; DDR, DNA damage response; HOMDEL, homologous deletion; MSI, microsatellite instability; ns, not significant; RFS, recurrence-free survival; RTK, receptor tyrosine kinase; TMB, tumor mutational burden; TRUNC, truncation.

Patients with either pathologic downstaging or durable complete clinical response had numerically higher median TMB (9.81 vs 4.25, p = 0.06), more strong binding neoantigens (233 vs 155.5, p = 0.11), and a higher fraction of APOBEC signature (0.53 vs 0.44, p = 0.47), though these differences were not statistically significant (Figure 2B). No genomic features were significantly associated with either EFS or RFS (Supplemental Figure S1).

Gene expression analysis

We next explored the gene expression profile of 25 samples from 18 patients to assess effects of treatment. Though there was low variance overall, the first principal component was correlated with timing of the biopsy, either pre- or post-treatment (r = 0.69, p ≤ 0.001), and paired samples from patients clustered together (Supplemental Figure S2A-B). Pre-treatment tumor samples also appeared to cluster by pathologic stage at the time of radical cystectomy, though this was not statistically significant (r = 0.4, p=0.25 (Supplemental Figures S2C-D). Differential gene expression analysis of pre-treatment samples (n = 10) and post-treatment samples (n = 15) independently exhibited upregulation of 38 and 28 genes and downregulation of 22 and 19 genes, respectively, all with 1.5x differential expression and adjusted p-value < 0.05. (Supplemental Figure S3A-B). Despite literature suggesting that DNA damage response and repair (DDR) gene mutations may be associated with benefit from immune checkpoint blockade (30), there was no association between DDR gene expression in pre-treatment tumors and response in our cohorts. Gene set enrichment analysis (GSEA) indicated changes in expression in several immune pathways in both pre-treatment and post-treatment samples (Supplemental Figure S3C-D). Specifically, we found upregulation of TNFA signaling and IL2 STAT5 signaling pathways in post-treatment samples suggesting a shift in immune content following treatment.

A previous study of neoadjuvant anti-PD-L1 therapy for MIBC with atezolizumab (the ABACUS trial), demonstrated an association between high immune signatures and pathologic response (32). Therefore, we considered immune content via immune deconvolution in our pre-treatment samples (Figure 3A). We found numerically longer RFS among cases with higher immune content (Figures 3B). When the same immune deconvolution and clusters were applied to gene expression data from PURE-01, a trial for MIBC of neoadjuvant anti-PD-1 therapy with pembrolizumab (33), pre-treatment tumors with hot immune clusters were associated with longer RFS (p = 0.022, Figure 3C, Supplemental Figure S4).

Figure 3.

(A) Heat map showing the distribution of immune cell types in pre-treatment tumor samples. (B-C) RFS by immune hot vs immune cold samples in (B) the current trial cohort and (C) PURE-01 as an external cohort.

MSK, Memorial Sloan Kettering; RFS, recurrence-free survival; TMB, tumor mutational burden.

Prior work on the PURE-01 and ABACUS cohorts identified an association between poor clinical outcomes on immune checkpoint blockade and a luminal-excluded/LumE subtype (24). We classified our cohort into subtypes accordingly (Supplemental Figure S5), and consistent with the prior findings, the only patient in our cohort with a luminal E subtype pre-treatment tumor experienced short EFS < 6 months. In contrast, the only patient in our cohort with a pre-treatment tumor with an immature basal subtype remained free of EFS events after > 3 years of follow-up.

Plasma cell-free DNA

Six patients in cohort 2 had sufficient plasma cell-free DNA yield from multiple timepoints for successful library preparation and longitudinal sequencing. As of last follow-up, 1 of the 6 experienced disease recurrence and 1 progressed prior to cystectomy. Dynamic changes in cell-free DNA tended to correspond with pathologic response at the time of surgery, wherein increasing variant allele frequencies (VAF) occurred in cases with pathologic non-response or progression prior to radical cystectomy (Supplemental Figure S6A). Patients without pathologic response also tended to have increasing cfDNA VAF post-operatively despite surgical resection (Supplemental Figure S6B). In another case, divergent changes in cfDNA VAF over time appeared to demonstrate divergent response in subclones to treatment (Supplemental Figure S6C). In some cases, pre-op and post-op cfDNA appeared to demonstrate superior prognostic value compared to pathologic response. For example, in one patient with ypT4N0 disease, post-op cfDNA was undetectable and the patient remained alive and free of recurrence as of last assessment at 18.9 months despite lack of pathologic response, though the patient did also receive adjuvant therapy (Supplemental Figure S6D). In the case of a separate patient with ypT2aN0 disease at time of surgery, pre-op cfDNA was undetectable and the patient remained alive and free of disease at 25.5 months (Supplemental Figure S6E). Similarly, one patient with ypT4aN0 disease at cystectomy who demonstrated decreased ctDNA levels before surgery remained free of cancer recurrence until death from unrelated causes at 15 months (Supplemental Figure S6F).

Analysis of peripheral blood mononuclear cells (PBMCs)

Pooling cohorts 1 and 2, exploratory flow cytometric analysis of PBMCs demonstrated that higher CTLA4 and ICOS positivity of CD4-positive T cells prior to radical cystectomy was associated with shorter EFS, with a HR of 3.60 associated with log percentage of CD4 T cells positive for CTLA4-positive cells (95% CI, 1.03–12.5, p = .044) and a HR of 5.67 associated with log percentage ICOS-positive cells (95% CI, 1.16–27.7, p = .032). Increase in the log percentage PD-1 positive CD8-positive T cells from baseline to midway through treatment was associated with a hazard ratio for death of 14.1 (95% CI, 2.49–79.7, p = .003). Longitudinal analysis of peripheral blood T cell receptor sequencing suggested greater variation in productive Simpson clonality and productive entropy between patients rather than within patients over time, and there was no clear difference in Simpson clonality or entropy trajectory over time between patients with treatment response versus non-response (Supplemental Figure S7).

DISCUSSION

In this investigator-initiated phase 2 trial of patients with MIBC ineligible for standard neoadjuvant cisplatin-based chemotherapy, neoadjuvant nivolumab alone plus RC-PLND was well-tolerated and achieved pathologic response and durable recurrence-free survival in a subset of patients. In contrast, addition of ipilimumab to nivolumab in cohort 2 caused TRAEs that prevented or delayed radical cystectomy in a significant portion of patients, without any apparent improvement in clinical efficacy. Cases of disease progression prior to surgery were observed in both treatment cohorts, indicating that immune checkpoint inhibition alone will be inadequate as a neoadjuvant strategy unless reliable predictive biomarkers are developed to guide patient selection. Concurrent biomarker analyses indicated potential prognostic and/or predictive markers in the context of neoadjuvant immune checkpoint blockade for MIBC, including circulating tumor DNA variant allele frequency, alterations of the nucleotide excision repair gene ERCC2 and the chromatin modifying gene NCOR1, and hot immune signature by gene expression profiling. The association of hot immune signatures in pre-treatment tumors and longer RFS remained consistent when examined in an external dataset of patients with MIBC treated with pembrolizumab. These findings are potentially informative for future studies aimed to predict response to immune checkpoint blockade in the neoadjuvant setting as well as bladder sparing approaches, acknowledging the limitations of our findings due to small patient numbers in the trial. On the other hand, we did not identify any biomarkers which correlated with progressive disease.

In our cohort of cisplatin ineligible patients receiving neoadjuvant therapy for MIBC, anti-PD-1 monotherapy in the form of nivolumab was well-tolerated and active, consistent with prior studies (32, 34). However, two patients (13%) progressed on nivolumab prior to radical cystectomy to a metastatic and incurable state, raising concerns about the safety of this treatment approach in unselected patients. Given recent acceptance of post-operative nivolumab as a standard adjuvant option for patients with muscle-invasive bladder cancer at high risk of recurrence, use of perioperative anti-PD-1 monotherapy is likely best suited to the post-operative setting (11), while the pre-operative setting should be reserved for more active neoadjuvant regimens—such as chemoimmunotherapy (3)—to minimize the potential for progression prior to potentially curative surgery. Unfortunately, addition of anti-CTLA4 blockade with ipilimumab to nivolumab in our study failed to produce significant improvement in clinical activity and resulted in significant toxicity. The low pathologic response rate observed with ipilimumab plus nivolumab in our trial differs from that of prior phase 2 studies of neoadjuvant anti-PD(L)1 plus anti-CTLA4 therapy for locoregional bladder cancer, which demonstrated pathologic complete response rates of 33–46% (35–37). Our study design and dosing strategies differed from other trials with anti-PD(L)1 plus anti-CTLA4 therapy, adopting a six-weekly ipilimumab schedule for Cohort 2 in an effort to reduce potential treatment-related toxicities, based on CheckMate-012 (38). These discrepant results may be attributable to dosing, given that 7 of 15 patients in our ipilimumab cohort received 50% of planned doses or less. Another potential explanation for the low response rates observed in our study may relate to the frailty of our cisplatin-ineligible patient population. Frailty is associated with distinct molecular features of peripheral blood immune cells that might impact immunotherapy efficacy (39) and poor performance status has been associated with inferior immune checkpoint inhibitor efficacy in tumor types ranging from urothelial carcinoma to lung cancer (40). Notably, our trial population was particularly frail; while many neoadjuvant trials for cisplatin ineligible MIBC allow patients who may be eligible for but decline cisplatin, our trial was specifically limited to patients who were ineligible to cisplatin due to comorbidities.

Some patients in our cohort who were unable to undergo radical cystectomy experienced durable complete clinical responses with immune checkpoint inhibition alone. Neoadjuvant studies such as the current trial remain a cornerstone of advancing precision medicine by facilitating correlative studies on matched pre- and post-treatment tissue for biomarker discovery and validation. While our correlative findings suggested potentially informative predictors of treatment efficacy, we were unable to identify biomarkers that could accurately and precisely predict exceptional responses. This result is consistent with prior studies of bladder cancer that have identified biomarkers such as PD-L1 and TMB as associated with response to immune checkpoint blockade but lacking in sufficient predictive precision for use in clinical decision making (35, 36, 41, 42). While neoadjuvant immune checkpoint blockade for patients who are ineligible for or refuse cisplatin-based chemotherapy shows promise across phase II trials for MIBC with pathologic complete response (pCR) rates of 31–42% (35, 36, 41, 42), superior biomarkers are clearly needed to optimize patient selection, especially for trials investigating bladder sparing approaches. Notably, all four patients in our cohort with loss of function alterations in the nucleotide excision repair gene ERCC2 had either pathologic downstaging to non-muscle invasive disease or a durable complete clinical response, though this finding requires validation in future studies. Alternatively, regimens combining immune checkpoint blockade with additional agents to achieve higher response rates may ultimately provide neoadjuvant options that are active enough for biomarker unselected, cisplatin-ineligible patient populations. For example, the highly active combination of the PD-1 inhibitor pembrolizumab with the antibody-drug conjugate enfortumab vedotin was recently shown to improve overall survival compared to cystectomy alone when given perioperatively to patients with muscle-invasive bladder cancer (6). Nonetheless, additional neoadjuvant options are still needed, since a proportion of patients are ineligible for both cisplatin and enfortumab vedotin due to comorbidities such as peripheral neuropathy and hepatic dysfunction (7).

An additional cohort for the current study is now enrolling patients with upper tract urothelial cancer for treatment with an alternative schedule of neoadjuvant ipilimumab and nivolumab (NCT03520491). Notably. the first stage of this cohort’s Simon optimal two-stage design showed pathologic downstaging in 67% of patients, including all 3 patients with germline mismatch repair alterations (43). Other potential future directions for this work include studies comparing neoadjuvant to adjuvant checkpoint blockade, versus neoadjuvant plus adjuvant combined. This is especially of interest given the established role of adjuvant nivolumab for resected MIBC at high risk of recurrence in conjunction with data in melanoma indicating that addition of neoadjuvant to adjuvant immune checkpoint blockade resulted in superior efficacy compared to adjuvant immune checkpoint blockade alone (11, 44).

Limitations of our sequential cohort design include a lack of randomization between treatment arms and small patient numbers, resulting in inability to formally compare efficacy and toxicities between treatment arms. In addition, a subset of patients did not have sufficient tissue available for biomarker analysis.

In conclusion, this phase 2 trial demonstrated that anti-PD-1 monotherapy with nivolumab prior to RC-PLND for patients with MIBC ineligible for standard cisplatin-based chemotherapy was well-tolerated and achieved durable recurrence-free survival in a subset of patients. However, combination neoadjuvant anti-PD-1 plus anti-CTLA4 therapy with nivolumab plus ipilimumab caused significant treatment-related toxicities that delayed or even prevented RC-PLND in a significant portion of cases, without apparent improvement in clinical efficacy. Given our findings, cisplatin-ineligible patients who are, by definition, relatively frail and tend to have greater difficulty tolerating treatment-related toxicities, may be best served by tolerable treatment approaches such as anti-PD-1 monotherapy rather than combination anti-PD-1 plus anti-CLTA4. Notably, some patients who were unable to undergo radical cystectomy experienced durable complete clinical responses. Further studies to identify patients who will achieve such exceptional responses to neoadjuvant immunotherapy are needed.

Translational Relevance.

Cisplatin-based neoadjuvant chemoimmunotherapy prior to radical cystectomy confers survival benefit for patients with muscle-invasive bladder cancer. However, many patients are ineligible for cisplatin due to comorbidities. We conducted a phase II trial of neoadjuvant nivolumab ± ipilimumab for cisplatin-ineligible patients with muscle-invasive bladder cancer. Nivolumab alone was well tolerated, while addition of ipilimumab caused toxicity that delayed curative intent cystectomy. Pathologic response rates were low in both cohorts, indicating limited anti-tumor efficacy of neoadjuvant nivolumab ± ipilimumab in biomarker unselected, cisplatin-ineligible patients. However, some patients experienced sustained clinical complete responses even without cystectomy, suggesting that neoadjuvant immunotherapy without cytotoxic chemotherapy may be a viable option for selected patients. Findings in the current trial, supported by analysis of data from the PURE-01 trial, indicate tumor infiltrating-immune cells in pre-treatment tissue may predict for durable recurrence-free survival after neoadjuvant immune checkpoint inhibitors.

Abbreviations list:

AE

adverse event

AMP

amplification

BCG

Bacillus Calmette-Guerin

cfDNA

cell-free DNA

CI

confidence interval

CR

complete response

DDR

DNA damage response and repair

EFS

event-free survival

FDR

false discovery rate

FFPE

formalin-fixed paraffin-embedded

GSEA

Gene Set Enrichment Analysis

HOMDEL

homologous deletion

HR

hazard ratio

IHC

immunohistochemistry

MIBC

muscle-invasive bladder cancer

MSI

microsatellite instability

MSI-H

microsatellite instability high

MSI-I

microsatellite instability indeterminate

MSK

Memorial Sloan Kettering

MSKCC

Memorial Sloan Kettering Cancer Center

MSS

microsatellite stable

No

number

ns

not significant

PBMC

peripheral blood mononuclear cell

PCA

principal component analysis

pCR

pathologic complete response

PCR

polymerase chain reaction

RC-PLND

radical cystectomy plus pelvic lymph node dissection

RFS

recurrence-free survival

RTK

receptor tyrosine kinase

TMB

tumor mutational burden

TRAE

treatment-related adverse event

TRUNC

truncation

TURBT

transurethral resection of bladder tumor

VAF

variant allele frequency

WES

whole exome sequencing